Research software for digital pathology

AdiFind: Whole-slide adipocyte segmentation and tumor-proximity analysis

Measure adipocyte size, spatial context, and tumor distance from gigapixel histology slides, with QuPath-ready outputs.

Download sample results Download sample .svs Contact

21,968

Slides processed

>75 million

Adipocytes measured

Overview

What AdiFind does

AdiFind segments adipocytes in whole-slide histology images, measures their size and spatial relationship to tumor, and exports workflow-ready results for downstream review and analysis.

Whole-slide adipocyte instance segmentation
Optional tumor-distance analysis
ROI-restricted processing
QuPath-ready annotation export
Annotated overlays and measurement tables

Outputs

What you get

Adipocyte-level measurements

CSV export with adipocyte IDs, areas, centroids, and optional tumor-distance fields.

Slide-level summaries

Structured JSON summary statistics for downstream analysis.

QuPath-ready annotations

GeoJSON export for direct review and integration in QuPath.

Annotated visual outputs

Whole-slide TIFF overlays with adipocyte detections and optional tumor-distance zones.

Featured

Annotated whole-slide output

AdiFind generates whole-slide overlays and export files ready for review, quantification, and downstream integration.

Annotated whole-slide image with tumor and adipose tissue context.

Integration

QuPath integration

AdiFind outputs integrate directly with existing QuPath workflows, including nuclear segmentation such as StarDist and spatial annotation pipelines.

Compatible with QuPath-based workflows.

Pipeline

How it works

Whole-slide image pyramid illustrating multi-resolution access from downscaled to full-resolution tissue.

Whole-slide context

AdiFind moves between slide-level context and full-resolution analysis using a multi-resolution whole-slide pyramid.

Downscaled 1X overview used for tumor detection and tissue guidance.

Guided tissue overview

A downscaled overview is used for tissue guidance and optional tumor analysis before high-resolution inference.

Full-resolution preprocessing showing raw, inverted, Sobel-enhanced, and adipocyte-detection stages.

Full-resolution detection

Full-resolution tissue windows are preprocessed to enhance adipocyte boundaries before instance detection.

Sliding-window inference with overlapping windows and reconciled adipocyte detections.

Tiled inference and reconciliation

AdiFind runs overlapping tiled inference across tissue regions and reconciles duplicate detections across tile boundaries.

Workflow Fit

Fits into existing pathology workflows

Desktop workflow

Desktop interface

Configure analysis settings and run AdiFind from a user-friendly desktop interface.

AdiFind is also accessible through a command-line interface, making it suitable for batch processing, scripted execution, and integration into existing computational pathology workflows.